Research On Signal Detection Technology Of Rolling Bearing Compound Failure

The rolling bearings are widely used in a variety of machines as standard parts,and rolling bearing fault is one of the main causes of machine failure. Fortunately,the vibration signal generated by a single failure of the rolling bearing has beenextensively studied and a series of powerful diagnostic methods have been made.However, in the course of the rolling bearing, there is not only have single fault,with the change of the operating environment and running time is increased, bearingoften have two or more faults that is named compound failure of the rolling bearing.Compound failure is not a simple superposition of multiple single faults, but thecoupling of multiple fault features, making the vibration signal becomescomplicated. Therefore, the study of compound fault signal detection technologiesbecome urgent to overcome.Based on the research on the signal characteristics of rolling bearings’compound fault,a method which based on the resonance underdetermined blindsource separation of sparse decomposition is proposed. Before detecting compoundfault signal, the first step is to separate compound fault signal into differentchannels, then use resonance sparse decomposition method for different channels offault signal to extract fault feature.Theoretical models of composite failure by two different kinds of rollingcompound fault are proposed: the outer ring and rolling elements，the inner ring andthe rolling elements. vibration signal modulation characteristics of rolling elementbearing surface with two faults are analyzed.Based on local mean decomposition of underdetermined blind sourceseparation method is proposed，which to solve compound fault single-channelrolling bearing vibration signal underdetermined blind source problem. Firstly,complex fault vibration signals were decomposed by local mean (LMD)decomposition, and contained different production function (PF), then use the blindsource separation method for PF components, and finally get with different faultindependent component signal characteristics.The impact resonance sparse decomposition method sparsity factors arestudied. By studying the correlation between the quality factor and the weight ratioof the coefficient of resonant sparse decomposition sparse extent, the most sparsedecomposition state optimization is maked.In order to verify the effectiveness of the proposed fault signal detectionmethods, experiments of widely used deep groove ball bearings are maked. Firstly,three different sets of rolling bearing compound fault are processed by electricaldischarge machining(EDM), and then use the signal detection method proposed in this paper to detect of bearing fault signal, the experimental results show that thismethod can not only separate the rolling bearing compound fault coupled signal intodifferent channels，but also can effectively extract the different the fault feature.Finally, this method compare with EMD and wavelet analysis, the results show thatwavelet analysis and EMD can extract the components of one fault, and the methodin this paper can not only fully extract all the fault feature, and get a higher a higherratio of signal to noise.